(Partially abridged from r.statistics.co)
ggplot2 is a data visualization package for the statistical programming language R. Created by Hadley Wickham in 2005, ggplot2 is an implementation of Leland Wilkinson’s Grammar of Graphics — a general scheme for data visualization which breaks up graphs into semantic components such as scales and layers. ggplot2 can serve as a replacement for the base graphics in R and contains a number of defaults for web and print display of common scales. Since 2005, ggplot2 has grown in use to become one of the most popular R packages. It is licensed under GNU GPL v2.
Nowadays ggplot2 is acknowledged as the most elegant and aesthetically pleasing graphics framework available in R.
In contrast to base R graphics, ggplot2 allows the user to add, remove or alter components in a plot at a high level of abstraction. This abstraction comes at a cost, with ggplot2 being slower than lattice graphics.
One potential limitation of base R graphics is the “pen-and-paper model” utilized to populate the plotting device. Graphical output from the interpreter is added directly to the plotting device or window rather than separately for each distinct element of a plot. In this respect it is similar to the lattice package, though Wickham argues ggplot2 inherits a more formal model of graphics from Wilkinson. As such, it allows for a high degree of modularity; the same underlying data can be transformed by many different scales or layers.
ggplot2 is usually loaded into R through tidyverse, a collection of R packages introduced by Hadley Wickham that share an underlying design philosophy, grammar, and data structures of tidy data (see Lab #1). The core packages are ggplot2, dplyr, tidyr, readr, purrr, tibble, stringr, and forcats, which provide functionality to model, transform, and visualize data. As of March 2021, the tidyverse package and some of its individual packages make up 5 out of the top 10 most downloaded R packages, and are the subject of multiple books and papers.
Moreover, a number of packages extending ggplot2 functionalities have appeared: among them, ggforce is one of the most relevant.
Let’s start loading the libraries
library(tidyverse)
The syntax for constructing ggplots could be puzzling if you are a
beginner or work primarily with base graphics. The main difference is
that, unlike base graphics, ggplot works with dataframes (tibbles) and
not individual vectors. All the data needed to make the plot is
typically be contained within the dataframe supplied to the
ggplot() call itself or can be supplied to respective
geoms. More on that later.
The second noticeable feature is that you can keep enhancing the plot
by adding more layers (and themes) to an existing plot created using the
ggplot() function.
Let’s initialize a basic ggplot based on the midwest
dataset.
# setup
options(scipen = 999) # turn off scientific notation like 1e+06
data("midwest", package = "ggplot2") # load the data
# midwest <- read.csv('http://goo.gl/G1K41K') # alt source
midwest
# init ggplot
ggplot(midwest, aes(x = area, y = poptotal)) # area and poptotal are columns in 'midwest'
A blank ggplot is drawn. Even though the x and y are specified, there are no points or lines in it. This is because ggplot doesn’t assume that you meant a scatterplot or a line chart to be drawn. I have only told ggplot what dataset to use and what columns should be used for X and Y axis. I haven’t explicitly asked it to draw any points.
Also note that the aes() function is used to specify the X and Y axes. That’s because any information that is part of the source dataframe has to be specified inside the aes() function.
Let’s make a scatterplot on top of the blank ggplot by adding points using a geom layer called geom_point.
ggplot(midwest, aes(x = area, y = poptotal)) + geom_point()
We got a basic scatterplot, where each point represents a county. However, it lacks some basic components such as the plot title, meaningful axis labels etc. Moreover, most of the points are concentrated on the bottom portion of the plot, which is not so nice. You will see how to rectify these in upcoming steps.
Like geom_point(), there are many such geom layers which we will see in a subsequent part in this lab. For now, let’s just add a smoothing layer using geom_smooth(method=‘lm’). Since the method is set as lm (short for linear model), it draws the line of best fit.
g <- ggplot(midwest, aes(x = area, y = poptotal)) + geom_point() +
geom_smooth(method = "lm") # set se=FALSE to turn off confidence bands
plot(g)
The line of best fit is in blue. Can you find out what other method options are available for geom_smooth? (note: see ?geom_smooth). You might have noticed that the majority of points still lie in the bottom of the chart which doesn’t really look nice.
So, let’s change the Y-axis limits to focus on the lower half.
The X and Y axis limits can be controlled in 2 ways.
This will change the lines of best fit or smoothing lines as compared to the original data.
This can be done by xlim() and ylim(). You can pass a numeric vector of length 2 (with max and min values) or just the max and min values itself.
g <- ggplot(midwest, aes(x = area, y = poptotal)) + geom_point() +
geom_smooth(method = "lm")
# delete the points outside the limits
g + xlim(c(0, 0.1)) + ylim(c(0, 1000000))
# equivalent: g + xlim(0, 0.1) + ylim(0, 1000000)
In this case, the chart was not built from scratch but rather was
built on top of g. This is because the previous plot was
stored as g, a ggplot object, which when called will reproduce
the original plot. Using ggplot, you can add more layers, themes and
other settings on top of this plot.
Did you notice that the line of best fit became more horizontal compared to the original plot? This is because, when using xlim() and ylim(), the points outside the specified range are deleted and will not be considered while drawing the line of best fit (using geom_smooth(method=‘lm’)). This feature might come in handy when you wish to know how the line of best fit would change when some extreme values (or outliers) are removed.
The other method is to change the X and Y axis limits by zooming in to the region of interest without deleting the points. This is done using coord_cartesian().
Let’s store this plot as g1.
g <- ggplot(midwest, aes(x = area, y = poptotal)) + geom_point() +
geom_smooth(method = "lm")
# Zoom in without deleting the points outside the limits.
# As a result, the line of best fit is the same as the
# original plot.
g1 <- g + coord_cartesian(xlim = c(0, 0.1), ylim = c(0, 1000000)) # zooms in
plot(g1)
Since all points were considered, the line of best fit did not change.
I have stored this as g1. Let’s add the plot title and labels for X and Y axis. This can be done in one go using the labs() function with title, x and y arguments. Another option is to use the ggtitle(), xlab() and ylab().
g <- ggplot(midwest, aes(x = area, y = poptotal)) + geom_point() +
geom_smooth(method = "lm")
g1 <- g + coord_cartesian(xlim = c(0, 0.1), ylim = c(0, 1000000)) # zooms in
# Add Title and Labels
g1 + labs(title = "Area Vs Population", subtitle = "From midwest dataset",
y = "Population", x = "Area", caption = "Midwest Demographics")
# equivalent:
g1 + ggtitle("Area Vs Population", subtitle = "From midwest dataset") +
xlab("Area") + ylab("Population")
Excellent! So we’ll see shortly the full function call.
We can change the aesthetics of a geom layer by modifying the respective geoms. Let’s change the color of the points and the line to a static value.
ggplot(midwest, aes(x=area, y=poptotal)) +
geom_point(col="steelblue", size=3) + # Set static color and size for points
geom_smooth(method="lm", col="firebrick") + # change the color of line
coord_cartesian(xlim=c(0, 0.1), ylim=c(0, 1000000)) +
labs(title="Area Vs Population", subtitle="From midwest dataset", y="Population", x="Area", caption="Midwest Demographics")
Suppose now we want the color to change based on another column in
the source dataset midwest: then, we must specify it inside
the aes() function.
gg <- ggplot(midwest, aes(x=area, y=poptotal)) +
geom_point(aes(col=state), size=3) + # Set color to vary based on state categories
geom_smooth(method="lm", col="firebrick", size=2) +
coord_cartesian(xlim=c(0, 0.1), ylim=c(0, 1000000)) +
labs(title="Area Vs Population", subtitle="From midwest dataset", y="Population", x="Area", caption="Midwest Demographics")
plot(gg) # or: print(gg)
Now each point is colored based on the state it belongs because of
aes(col=state). Not just color, but size, shape, stroke
(thickness of boundary) and fill (fill color) can be used to
discriminate groupings.
As a bonus, the legend is added automatically. If needed, it can be removed by setting the legend.position to None from within a theme() function:
gg + theme(legend.position = "None") # remove legend
Also, you can change the color palette entirely:
gg + scale_colour_brewer(palette = "Set1") # change color palette
More of such palettes can be found in the RColorBrewer package:
library(RColorBrewer)
head(brewer.pal.info, 10) # show 10 palettes
Now let’s see how to change the X and Y axis text and its location. This involves two aspects: breaks and labels.
The breaks should be of the same scale as the X axis variable. Note that I am using scale_x_continuous because, the X axis variable is a continuous variable. Had it been a date variable, scale_x_date could be used. Like scale_x_continuous() an equivalent scale_y_continuous() is available for Y axis.
# Base plot
gg <- ggplot(midwest, aes(x=area, y=poptotal)) +
geom_point(aes(col=state), size=3) + # Set color to vary based on state categories.
geom_smooth(method="lm", col="firebrick", size=2) +
coord_cartesian(xlim=c(0, 0.1), ylim=c(0, 1000000)) +
labs(title="Area Vs Population", subtitle="From midwest dataset", y="Population", x="Area", caption="Midwest Demographics")
# Change breaks
gg + scale_x_continuous(breaks=seq(0, 0.1, 0.01))
You can optionally change the labels at the axis ticks with the
argument labels=*, which takes a vector of the same length
as breaks. Let me demonstrate by setting the labels to
alphabets from a to k (though there is no meaning to it in this
context).
# Base Plot
gg <- ggplot(midwest, aes(x=area, y=poptotal)) +
geom_point(aes(col=state), size=3) + # Set color to vary based on state categories.
geom_smooth(method="lm", col="firebrick", size=2) +
coord_cartesian(xlim=c(0, 0.1), ylim=c(0, 1000000)) +
labs(title="Area Vs Population", subtitle="From midwest dataset", y="Population", x="Area", caption="Midwest Demographics")
# Change breaks + label
gg + scale_x_continuous(breaks=seq(0, 0.1, 0.01), labels = letters[1:11])
Let’s set the breaks for Y axis text as well and format the X and Y axis labels. I have used two methods for formatting labels:
area as % in
the example belowpopulation to 1K scaleUse whichever method feels convenient.
# Base Plot
gg <- ggplot(midwest, aes(x=area, y=poptotal)) +
geom_point(aes(col=state), size=3) + # Set color to vary based on state categories.
geom_smooth(method="lm", col="firebrick", size=2) +
coord_cartesian(xlim=c(0, 0.1), ylim=c(0, 1000000)) +
labs(title="Area Vs Population", subtitle="From midwest dataset", y="Population", x="Area", caption="Midwest Demographics")
# Change Axis Texts
gg + scale_x_continuous(breaks=seq(0, 0.1, 0.01), labels = sprintf("%1.2f%%", seq(0, 0.1, 0.01))) +
scale_y_continuous(breaks=seq(0, 1000000, 200000), labels = function(x){paste0(x/1000, 'K')})
Finally, instead of changing the theme components individually (which I discuss in detail in a subsequent part), we can change the entire theme itself using pre-built themes. The help page ?theme_bw shows all the available built-in themes.
Again, this is commonly done in a couple of ways.
# Base plot
gg <- ggplot(midwest, aes(x=area, y=poptotal)) +
geom_point(aes(col=state), size=3) + # Set color to vary based on state categories.
geom_smooth(method="lm", col="firebrick", size=2) +
coord_cartesian(xlim=c(0, 0.1), ylim=c(0, 1000000)) +
labs(title="Area Vs Population", subtitle="From midwest dataset", y="Population", x="Area", caption="Midwest Demographics")
gg <- gg + scale_x_continuous(breaks=seq(0, 0.1, 0.01))
# method 1: Using theme_set()
theme_set(theme_classic()) # not run
gg
# method 2: Adding theme as a layer itself
gg + theme_bw() + labs(subtitle = "BW Theme")
gg + theme_classic() + labs(subtitle = "Classic Theme")
For more customized and fancy themes have a look at the ggthemes package and the ggthemr package.
Let’s begin with a scatterplot of Population against Area from midwest dataset. The point’s color and size vary based on state (categorical) and popdensity (continuous) columns respectively. The below plot has the essential components such as the title, axis labels and legend setup nicely. But how to modify the looks?
Most of the requirements related to look and feel can be achieved using the theme() function, which accepts a large number of arguments. Type ?theme in the R console and see for yourself.
# if you modified the theme by theme_set() previously:
# theme_set(theme_bw())
# Add plot components --------------------------------
gg <- ggplot(midwest, aes(x = area, y = poptotal)) + geom_point(aes(col = state,
size = popdensity)) + geom_smooth(method = "loess", se = F) +
xlim(c(0, 0.1)) + ylim(c(0, 500000)) + labs(title = "Area Vs Population",
y = "Population", x = "Area", caption = "Source: midwest")
# Call plot ------------------------------------------
plot(gg)
The arguments passed to theme() components require to be set using special element_type() functions. They are of 4 major types.
More on this follows in upcoming discussion.
Let’s discuss a number of tasks related to changing the plot output, starting with modifying the title and axis texts.
Plot and axis titles and the axis text are part of the plot’s theme. Therefore, it can be modified using the theme() function. The theme() function accepts one of the four element_type() functions mentioned above as arguments. Since the plot and axis titles are textual components, element_text() is used to modify them.
Below, I have changed the font size, color, face, and line height. The axis text can be rotated by changing the angle.
# Base Plot
gg <- ggplot(midwest, aes(x=area, y=poptotal)) +
geom_point(aes(col=state, size=popdensity)) +
geom_smooth(method="loess", se=FALSE) + xlim(c(0, 0.1)) + ylim(c(0, 500000)) +
labs(title="Area Vs Population", y="Population", x="Area", caption="Source: midwest")
# Modify theme components -------------------------------------------
gg + theme(plot.title=element_text(size=20,
face="bold",
family="Roboto",
color="tomato",
hjust=0.5,
lineheight=1.2), # title
plot.subtitle=element_text(size=15,
family="Roboto",
face="bold",
hjust=0.5), # subtitle
plot.caption=element_text(size=15), # caption
axis.title.x=element_text(vjust=0,
size=15), # X axis title
axis.title.y=element_text(size=15), # Y axis title
axis.text.x=element_text(size=10,
angle = 30,
vjust=.5), # X axis text
axis.text.y=element_text(size=10)) # Y axis text
Above example covers some of the frequently used theme modifications and the actual list is too long. So ?theme is the first place you want to look at if you want to change the look and feel of any component.
Whenever your plot’s geom (like points, lines, bars,
etc) is set to change the aesthetics (fill, size, col, shape or stroke)
based on another column, as in
geom_point(aes(col=state, size=popdensity)), a legend is
automatically drawn.
If you are creating a geom where the aesthetics are
static, a legend is not drawn by default. In such cases you might want
to create your own legend manually. The below examples are for cases
where you have the legend created automatically.
Let’s now change the legend title. We have two legends, one each for color and size. The size is based on a continuous variable while the color is based on a categorical (discrete) variable.
There are 3 ways to change the legend title.
# Base Plot
gg <- ggplot(midwest, aes(x = area, y = poptotal)) + geom_point(aes(col = state,
size = popdensity)) + geom_smooth(method = "loess", se = F) +
xlim(c(0, 0.1)) + ylim(c(0, 500000)) + labs(title = "Area Vs Population",
y = "Population", x = "Area", caption = "Source: midwest")
gg + labs(color = "State", size = "Density") # modify legend title
# Base Plot
gg <- ggplot(midwest, aes(x = area, y = poptotal)) + geom_point(aes(col = state,
size = popdensity)) + geom_smooth(method = "loess", se = F) +
xlim(c(0, 0.1)) + ylim(c(0, 500000)) + labs(title = "Area Vs Population",
y = "Population", x = "Area", caption = "Source: midwest")
gg + guides(color = guide_legend("State"), size = guide_legend("Density")) # modify legend title
scale_<aesthetic>_<vartype>()
formatThe format of scale_<aesthetic>_<vartype>()
allows you to turn off legend for one particular aesthetic, leaving the
rest in place. This can be done just by setting
guide=FALSE. For example, if the legend is for size of
points based on a continuous variable, then
scale_size_continuous() would be the right function to use.
# Base Plot
gg <- ggplot(midwest, aes(x = area, y = poptotal)) + geom_point(aes(col = state,
size = popdensity)) + geom_smooth(method = "loess", se = F) +
xlim(c(0, 0.1)) + ylim(c(0, 500000)) + labs(title = "Area Vs Population",
y = "Population", x = "Area", caption = "Source: midwest")
# Modify Legend
gg + scale_color_discrete(name = "State") + scale_size_continuous(name = "Density",
guide = FALSE) # turn off legend for size
This can be done using the respective scale_aesthetic_manual() function. The new legend labels are supplied as a character vector to the labels argument. If you want to change the color of the categories, it can be assigned to the values argument as shown in below example.
gg <- ggplot(midwest, aes(x = area, y = poptotal)) + geom_point(aes(col = state,
size = popdensity)) + geom_smooth(method = "loess", se = F) +
xlim(c(0, 0.1)) + ylim(c(0, 500000)) + labs(title = "Area Vs Population",
y = "Population", x = "Area", caption = "Source: midwest")
gg + scale_color_manual(name = "State", labels = c("Illinois",
"Indiana", "Michigan", "Ohio", "Wisconsin"), values = c(IL = "blue",
IN = "red", MI = "green", OH = "brown", WI = "orange"))
In case you want to show the legend for color (State) before size (Density), it can be done with the guides() function. The order of the legend has to be set as desired.
If you want to change the position of the labels inside the legend, set it in the required order as seen in previous example.
# Base Plot
gg <- ggplot(midwest, aes(x = area, y = poptotal)) + geom_point(aes(col = state,
size = popdensity)) + geom_smooth(method = "loess", se = F) +
xlim(c(0, 0.1)) + ylim(c(0, 500000)) + labs(title = "Area Vs Population",
y = "Population", x = "Area", caption = "Source: midwest")
gg + guides(colour = guide_legend(order = 1), size = guide_legend(order = 2))
The styling of legend title, text, key and the guide can also be adjusted. The legend’s key is a figure-like element, so it has to be set using element_rect() function.
# Base Plot
gg <- ggplot(midwest, aes(x = area, y = poptotal)) + geom_point(aes(col = state,
size = popdensity)) + geom_smooth(method = "loess", se = F) +
xlim(c(0, 0.1)) + ylim(c(0, 500000)) + labs(title = "Area Vs Population",
y = "Population", x = "Area", caption = "Source: midwest")
gg + theme(legend.title = element_text(size = 12, color = "firebrick"),
legend.text = element_text(size = 10), legend.key = element_rect(fill = "springgreen")) +
guides(colour = guide_legend(override.aes = list(size = 2,
stroke = 1.5)))
The legend’s position inside the plot is an aspect of the theme. So it can be modified using the theme() function. If you want to place the legend inside the plot, you can additionally control the hinge point of the legend using legend.justification.
The legend.position is the x and y axis position in chart area, where (0,0) is bottom left of the chart and (1,1) is top right. However, there are useful string presets such as “left”/“right” and “top”/“bottom”. Likewise, legend.justification refers to the hinge point inside the legend.
# Base Plot
gg <- ggplot(midwest, aes(x = area, y = poptotal)) + geom_point(aes(col = state,
size = popdensity)) + geom_smooth(method = "loess", se = F) +
xlim(c(0, 0.1)) + ylim(c(0, 500000)) + labs(title = "Area Vs Population",
y = "Population", x = "Area", caption = "Source: midwest")
# No legend
# --------------------------------------------------
gg + theme(legend.position = "None") + labs(subtitle = "No Legend")
# Legend to the left
# -----------------------------------------
gg + theme(legend.position = "left") + labs(subtitle = "Legend on the Left")
# legend at the bottom and horizontal
# ------------------------
gg + theme(legend.position = "bottom", legend.box = "horizontal") +
labs(subtitle = "Legend at Bottom")
# legend at bottom-right, inside the plot
# --------------------
gg + theme(legend.title = element_text(size = 12, color = "salmon",
face = "bold"), legend.justification = c(1, 0), legend.position = c(0.95,
0.05), legend.background = element_blank(), legend.key = element_blank()) +
labs(subtitle = "Legend: Bottom-Right Inside the Plot")
# legend at top-left, inside the plot
# -------------------------
gg + theme(legend.title = element_text(size = 12, color = "salmon",
face = "bold"), legend.justification = c(0, 1), legend.position = c(0.05,
0.95), legend.background = element_blank(), legend.key = element_blank()) +
labs(subtitle = "Legend: Top-Left Inside the Plot")
Let’s try adding some text. We will add text to only those counties that have population greater than 300K. In order to achieve this, I create another subsetted dataframe (midwest_sub) that contains only the qualifying counties.
Then, draw the geom_text and geom_label with this new dataframe as the data source. This will ensure that labels (geom_label) are added only for the points contained in the new dataframe.
# Filter required rows.
midwest_sub <- midwest %>%
dplyr::filter(poptotal > 300000)
midwest_sub$large_county <- ifelse(midwest_sub$poptotal > 300000,
midwest_sub$county, "")
# Base Plot
gg <- ggplot(midwest, aes(x = area, y = poptotal)) + geom_point(aes(col = state,
size = popdensity)) + geom_smooth(method = "loess", se = F) +
xlim(c(0, 0.1)) + ylim(c(0, 500000)) + labs(title = "Area Vs Population",
y = "Population", x = "Area", caption = "Source: midwest")
# Plot text and label
# ------------------------------------------------------
# here we use midwest_sub as the data source
gg + geom_text(aes(label = large_county), size = 2, data = midwest_sub) +
labs(subtitle = "With ggplot2::geom_text") + theme(legend.position = "None") # text
gg + geom_label(aes(label = large_county), size = 2, data = midwest_sub,
alpha = 0.25) + labs(subtitle = "With ggplot2::geom_label") +
theme(legend.position = "None") # label
# Plot text and label that REPELS each other (using ggrepel
# pkg) ------------
library(ggrepel)
gg + geom_text_repel(aes(label = large_county), size = 2, data = midwest_sub) +
labs(subtitle = "With ggrepel::geom_text_repel") + theme(legend.position = "None") # text
gg + geom_label_repel(aes(label = large_county), size = 2, data = midwest_sub) +
labs(subtitle = "With ggrepel::geom_label_repel") + theme(legend.position = "None") # label
Since the label is looked up from a different dataframe, we need to
set it through the data argument.
Let’s see how to add annotation to any specific point of the chart.
It can be done with the annotation_custom() function which
takes in a grob (“grid graphical object”) as the argument. So,
let’s create a grob that holds the text you want to display using the
grid package.
# Base Plot
gg <- ggplot(midwest, aes(x = area, y = poptotal)) + geom_point(aes(col = state,
size = popdensity)) + geom_smooth(method = "loess", se = F) +
xlim(c(0, 0.1)) + ylim(c(0, 500000)) + labs(title = "Area Vs Population",
y = "Population", x = "Area", caption = "Source: midwest")
# Define and add annotation
# -------------------------------------
library(grid)
my_text <- "This text is at x=0.7 and y=0.8!"
gg + annotation_custom(grob = grid.text(my_text, x = 0.6, y = 0.8,
gp = gpar(col = "firebrick", fontsize = 14, fontface = "bold")))
Easy: just add coord_flip().
# Base Plot
gg <- ggplot(midwest, aes(x = area, y = poptotal)) + geom_point(aes(col = state,
size = popdensity)) + geom_smooth(method = "loess", se = F) +
xlim(c(0, 0.1)) + ylim(c(0, 500000)) + labs(title = "Area Vs Population",
y = "Population", x = "Area", caption = "Source: midwest",
subtitle = "X and Y axis Flipped") + theme(legend.position = "None")
print(gg)
# Flip the X and Y axis
# -------------------------------------------------
gg + coord_flip()
This is quite simple. Use scale_x_reverse() for X axis and scale_y_reverse() for Y axis.
# Base Plot
gg <- ggplot(midwest, aes(x = area, y = poptotal)) + geom_point(aes(col = state,
size = popdensity)) + geom_smooth(method = "loess", se = F) +
labs(title = "Area Vs Population", y = "Population", x = "Area",
caption = "Source: midwest", subtitle = "Axis Scales Reversed") +
theme(legend.position = "None")
print(gg)
# Reverse the X and Y Axis ---------------------------
gg + scale_x_reverse() + scale_y_reverse()
Let’s use the mpg dataset for this one. It is available in the ggplot2 package, or you can import it from this link.
data(mpg, package = "ggplot2") # load data
# mpg <- read.csv('http://goo.gl/uEeRGu') # alt data source
mpg
g <- ggplot(mpg, aes(x = displ, y = hwy)) + geom_point() + labs(title = "hwy vs displ",
caption = "Source: mpg") + geom_smooth(method = "lm", se = FALSE) +
theme_bw()
plot(g)
We have a simple chart of highway mileage (hwy) against the engine displacement (displ) for the whole dataset. But what if you want to study how this relationship varies for different classes of vehicles?
The facet_wrap() is used to break down a large plot into
multiple small plots for individual categories. It takes an R
formula as the main argument. The items to the left of
~ form the rows while those to the right form the
columns.
By default, all the plots share the same scale in both X and Y axis.
You can set them free by setting scales='free' but this way
it could be harder to compare between groups.
# Base Plot
g <- ggplot(mpg, aes(x = displ, y = hwy)) + geom_point() + geom_smooth(method = "lm",
se = FALSE) + theme_bw()
# Facet wrap with common scales
g + facet_wrap(~class, nrow = 3) + labs(title = "hwy vs displ",
caption = "Source: mpg", subtitle = "ggplot2 - Faceting - Multiple plots in one figure") # Shared scales
# Facet wrap with free scales
g + facet_wrap(~class, scales = "free") + labs(title = "hwy vs displ",
caption = "Source: mpg", subtitle = "ggplot2 - Faceting - Multiple plots in one figure with free scales") # Scales free
So, what do you infer from this? For one, most 2seater cars have higher engine displacement while the minivan and compact vehicles are on the lower side. This is evident from where the points are placed along the X-axis.
Also, the highway mileage drops across all segments as the engine displacement increases. This drop seems more pronounced in compact and subcompact vehicles.
The headings of the middle and bottom rows take up significant space. The facet_grid() would get rid of it and give more area to the charts. The main difference with facet_grid is that it is not possible to choose the number of rows and columns in the grid.
Let’s create a grid to see how it varies with manufacturer.
# Base Plot
g <- ggplot(mpg, aes(x = displ, y = hwy)) + geom_point() + labs(title = "hwy vs displ",
caption = "Source: mpg", subtitle = "ggplot2 - Faceting - Multiple plots in one figure") +
geom_smooth(method = "lm", se = FALSE) + theme_bw()
# Add Facet Grid
g1 <- g + facet_grid(manufacturer ~ class) # manufacturer in rows and class in columns
plot(g1)
Let’s make one more to vary by cylinder.
# Base Plot
g <- ggplot(mpg, aes(x = displ, y = hwy)) + geom_point() + geom_smooth(method = "lm",
se = FALSE) + labs(title = "hwy vs displ", caption = "Source: mpg",
subtitle = "Ggplot2 - Facet Grid - Multiple plots in one figure") +
theme_bw() # apply bw theme
# Add Facet Grid
g2 <- g + facet_grid(cyl ~ class) # cyl in rows and class in columns.
plot(g2)
It is possible to layout both these charts in the sample panel. I prefer the gridExtra() package for this.
# Draw Multiple plots in same figure.
library(gridExtra)
gridExtra::grid.arrange(g1, g2, ncol = 2)
The background is a figure-like object, so we use
element_rect inside a call to theme():
# Base Plot
g <- ggplot(mpg, aes(x=displ, y=hwy)) +
geom_point() +
geom_smooth(method="lm", se=FALSE) +
theme_bw() # apply bw theme
print(g)
# Change Plot Background elements -----------------------------------
g + theme(panel.background = element_rect(fill = 'khaki'),
panel.grid.major = element_line(colour = "burlywood", size=1.5),
panel.grid.minor = element_line(colour = "tomato",
size=.25,
linetype = "dashed"),
panel.border = element_blank(),
axis.line.x = element_line(colour = "darkorange",
size=1.5,
lineend = "butt"),
axis.line.y = element_line(colour = "darkorange",
size=1.5)) +
labs(title="Modified Background",
subtitle="How to Change Major and Minor grid, Axis Lines, No Border")
# Change Plot Margins -----------------------------------------------
g + theme(plot.background=element_rect(fill="salmon"),
plot.margin = unit(c(2, 2, 1, 1), "cm")) + # top, right, bottom, left
labs(title="Modified Background", subtitle="How to Change Plot Margin")
# Base Plot
g <- ggplot(mpg, aes(x = displ, y = hwy)) + geom_point() + geom_smooth(method = "lm",
se = FALSE) + theme_bw()
g + theme(panel.grid.major = element_blank(), panel.grid.minor = element_blank(),
panel.border = element_blank(), axis.title = element_blank(),
axis.text = element_blank(), axis.ticks = element_blank()) +
labs(title = "Modified Background", subtitle = "How to remove major and minor axis grid, border, axis title, text and ticks")
library(grid)
library(png)
library(RCurl) # for downloading URLs
img <- png::readPNG(getURLContent("https://upload.wikimedia.org/wikipedia/commons/c/c1/Rlogo.png"))
g_pic <- rasterGrob(img, interpolate = TRUE)
# Base Plot
g <- ggplot(mpg, aes(x = displ, y = hwy)) + geom_point() + geom_smooth(method = "lm",
se = FALSE) + theme_bw()
g + theme(panel.grid.major = element_blank(), panel.grid.minor = element_blank(),
plot.title = element_text(size = rel(1.5), face = "bold"),
axis.ticks = element_blank()) + annotation_custom(g_pic,
xmin = 5, xmax = 7, ymin = 30, ymax = 45)